MTN South Africa is increasingly turning to artificial intelligence (AI) to manage its network, including to track down people stealing revenue and analyse traffic patterns.
MTN SA CEO Ferdi Moolman says the operator’s adoption of AI has accelerated as the technology finds practical applications across the business. One application is revenue assurance, where MTN SA is using AI to identify potential revenue leakage, he says.
Moolman points to zero-rated sites, including university websites, as an example, where fraudsters use a technique known as tunnelling to redirect their traffic to other websites that would usually result in mobile operators being able to charge for data use.
Zero-rating dates back to South Africa’s COVID-19 response, when the Independent Communications Authority of South Africa (ICASA) required mobile network operators to provide data-free access to all health and educational websites approved by government.
Subsequently, as part of the 2022 spectrum auction licence requirements, mobile operators must zero-rate qualifying online content from public benefit organisations, including government websites, by 15 January 2027.
Catching the rats
In a November 2021 submission to ICASA, MTN said it had “experienced high levels of data tunnelling as a result” of the COVID-19 requirement to allow open access to certain websites.
“Information is freely available on how to tunnel; for example, we have discovered tutorials that are freely available on YouTube and Telegram,” it says in its submission.
Moolman explains that, in practice, it’s very difficult to zero-rate a site because of the risk of hackers using tunnelling to bypass the revenue system. MTN has been putting measures in place to detect such activity, but he says the operator has again identified abuse of zero-rated sites.
MTN SA is now using AI to analyse behaviour patterns and identify where tunnelling is most likely taking place, says Moolman, who took up the CEO post last November.
This allows MTN SA to focus on a particular site where suspicious activity has been identified rather than affecting all zero-rated sites, he says.
AI is also being used to analyse how people use MTN’s network, with data being processed anonymously in what Moolman calls “a clean room”.
In the zone
The analysis gives MTN insight into behaviour that would be difficult to derive manually from the volume of data generated across its network, says Moolman. This includes understanding which online services are consuming data and where network traffic is being generated.
“It gives us a view in terms of which sites people are going to, how much [data] YouTube is consuming, how much is Facebook consuming,” Moolman says.
This analysis, he explains, can be extended to geographies. He cites Sandton as an example, where a combination of residential and office activity makes analysing network usage more complex.
AI also allows MTN to revisit concepts similar to those used by its former MTN Zone offering, but for the considerably more complex data environment, Moolman says. MTN Zone, launched in 2008, used dynamic per-second tariffs to offer discounts based on a customer’s location, time of day and demand on the network.
Applying a similar approach to data is more complicated because MTN’s network carries traffic across several generations of mobile technology. “It’s very difficult to do this with data, because you’ve got 3G, 4G, 5G data… You need AI to be able to do this,” he says.
Share
